Strongly Chordal Digraphs: Orderability & Algorithms

Updated 12 July 2026

Strongly chordal digraphs are defined by a symmetric Γ‐free ordering of their adjacency matrices, enforcing a global nesting of vertex neighborhoods.
They generalize strongly chordal and chordal bipartite graphs by unifying matrix orderability and strong vertex orderings, with applications in domination and elimination algorithms.
Recent studies offer linear-time recognition for key subclasses like tournaments and balanced digraphs, though a general polynomial-time recognition algorithm remains open.

Strongly chordal digraphs are digraphs whose adjacency matrices admit a symmetric $\Gamma$ -free ordering, where

$\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$

Equivalently, they are digraphs that admit a strong ordering of the vertex set. The notion was introduced as a common framework extending strongly chordal graphs and chordal bipartite graphs, while remaining strictly inside a broader chordal-digraph landscape. Its modern development is driven by matrix orderability, elimination orderings, obstruction sets, and specialized recognition algorithms; by contrast, general recognition for arbitrary digraphs with possible loops remains open (Hell et al., 2019, Hell et al., 23 Sep 2025).

1. Definition and matrix formulation

The defining property is a simultaneous row-and-column ordering of the adjacency matrix. For a square $0,1$-matrix $M$ , a symmetric $\Gamma$ -free ordering means that the same permutation is applied to rows and columns and the resulting matrix contains no submatrix equal to $\Gamma$ . A digraph $D$ is strongly chordal exactly when its adjacency matrix $M(D)$ has such an ordering (Hell et al., 2019).

This matrix condition has a purely combinatorial reformulation. An ordering $v_1,\dots,v_n$ of $V(D)$ is a strong ordering if, for all $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$0 and $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$1,

$\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$2

A digraph is strongly chordal if and only if it has a strong ordering. In the later formulation of the theory, the same condition is expressed with arc notation: $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$3 This ordering criterion is the direct digraph analogue of the strong ordering characterization of undirected strongly chordal graphs (Hell et al., 23 Sep 2025).

The matrix perspective is not merely a reformulation. It identifies strong chordality with a global orderability constraint rather than with a local cycle prohibition. The defining feature is the global nesting of neighborhoods forced by the forbidden $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$4 pattern under a single row/column order. This is substantially stronger than allowing independent row and column permutations, which leads instead to the classical notion of total balance (Hell et al., 2019).

2. Relation to classical graph classes and to chordal digraphs

Strongly chordal digraphs were introduced to unify two classical classes. First, if the digraph is symmetric and reflexive, then the digraph notion reduces to the usual notion of a strongly chordal graph: a reflexive graph is strongly chordal exactly when its adjacency matrix has a symmetric $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$5-free ordering. Second, chordal bipartite graphs also fit the framework, because the symmetric $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$6-free ordering condition on the digraph adjacency matrix specializes to the usual $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$7-free ordering of the biadjacency matrix in the bipartite setting (Hell et al., 2019, Hell et al., 23 Sep 2025).

Within directed graph theory, strong chordality sits inside a larger class of chordal digraphs. In the framework of the foundational paper, a digraph is chordal if it has a simplicial ordering, where a vertex $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$8 is simplicial if for every $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$9 and $0,1$0, the arc $0,1$1 is present. A strong ordering is stricter than a simplicial ordering, and indeed every strong ordering is a simple ordering, every simple ordering is a simplicial ordering, and therefore every strongly chordal digraph is chordal (Hell et al., 2019).

The converse fails. Strong chordality is more restrictive because it excludes the $0,1$2 pattern globally, not only at the current elimination vertex. This difference is visible already in tournaments: simple orderings and strong orderings do not coincide for general digraphs or tournaments (Hell et al., 2019).

A separate terminological issue arises because another line of work uses “chordal directed graph” for a digraph with no induced directed cycle of length at least $0,1$3. That cycle-based condition is much weaker. In particular, there exist digraphs with no $0,1$4 and no induced directed cycle of length at least $0,1$5 whose dichromatic number is arbitrarily large, so this broader class is not $0,1$6-bounded (Aboulker et al., 2022). This shows that strong chordality cannot be reduced to the prohibition of long induced directed cycles.

3. Structural properties, simple vertices, and obstruction theory

Several structural lemmas organize the subject. A vertex $0,1$7 is simplicial if either it is irreflexive, or for all $0,1$8 and $0,1$9, one has $M$ 0. A vertex is simple if it is simplicial and additionally the out-neighborhoods of any two in-neighbors are comparable by inclusion, while the in-neighborhoods of any two out-neighbors are comparable by inclusion. A simple ordering is one in which each vertex is simple in the remaining induced subdigraph. If no vertex of a digraph is simple, then the digraph is not strongly chordal (Hell et al., 23 Sep 2025).

The theory also isolates local obstructions. In an irreflexive digraph, a peak is a vertex $M$ 1 for which there exist $M$ 2 such that $M$ 3, $M$ 4, and $M$ 5. A peak cannot be the last vertex of a simple ordering; hence, if every vertex is a peak, the digraph is not strongly chordal. Another matrix-theoretic constraint is that if a digraph has a simple ordering, then its adjacency matrix is totally balanced, but the converse fails in general, even among irreflexive tournaments (Hell et al., 23 Sep 2025).

The symmetric case yields the fullest general characterization. For a graph with possible loops, the following are equivalent: the graph is strongly chordal; its adjacency matrix is totally balanced; every induced subgraph has a simple vertex; and the graph has a simple ordering. This extends Farber’s theorem from reflexive strongly chordal graphs to graphs with possible loops (Hell et al., 2019).

That symmetric theory admits both walk and obstruction descriptions. A graph with possible loops is strongly chordal if and only if every even closed walk of length at least $M$ 6 has a strong chord. Equivalently, it is strongly chordal if and only if it avoids the nine families $M$ 7, consisting of reflexive cycles of length at least $M$ 8, irreflexive cycles of length other than $M$ 9, several looped cycle-and-fan configurations, weak trampolines, and paths of length at least $\Gamma$ 0 whose endpoints are both reflexive (Hell et al., 2019).

4. Exact classifications in major subclasses

Later work develops complete descriptions for several important subclasses of digraphs with possible loops (Hell et al., 23 Sep 2025).

Class	Exact characterization	Recognition
Irreflexive tournaments	Obtained from a transitive tournament by reversing at most one arc; equivalently $\Gamma$ 1-free	$\Gamma$ 2, certifying
Tournaments with possible loops	Obtained from a transitive tournament by reversing at most one arc with at least one irreflexive endvertex; equivalently $\Gamma$ 3-free	$\Gamma$ 4, certifying
Reflexive multipartite tournaments	$\Gamma$ 5-free transitive orientations of complete split graphs	$\Gamma$ 6, certifying
Irreflexive bipartite tournaments	$\Gamma$ 7-free; equivalently one one-way bigraph has at most two nontrivial bipartite-chain components	$\Gamma$ 8
Balanced digraphs	Strongly chordal iff each layer $\Gamma$ 9 is a chordal bigraph; equivalently fence-free	Polynomial-time; linear-time in the later treatment

For irreflexive tournaments, the structure is especially rigid. Such a tournament is strongly chordal if and only if it is obtained from a transitive tournament by reversing at most one arc; equivalently, it does not contain two arc-disjoint directed triangles; equivalently, it is $\Gamma$ 0-free, where $\Gamma$ 1 is the family of six minimal forbidden induced subdigraphs. In the strongly connected case, the only possibility is $\Gamma$ 2, obtained from the transitive tournament $\Gamma$ 3 by reversing the arc from the unique source to the unique sink (Hell et al., 23 Sep 2025).

For tournaments with possible loops, the class is again close to transitivity. A tournament with possible loops is strongly chordal if and only if it is obtained from a transitive tournament by reversing at most one arc with at least one irreflexive endvertex; equivalently, it is $\Gamma$ 4-free; equivalently, it is $\Gamma$ 5-free, its irreflexive part is strongly chordal, and its reflexive vertices induce a transitive tournament (Hell et al., 23 Sep 2025).

Reflexive multipartite tournaments satisfy an even sharper global constraint. If such a digraph is strongly chordal, then its underlying graph is a complete split graph $\Gamma$ 6, where $\Gamma$ 7 induces a transitive tournament, $\Gamma$ 8 is an independent set, and every vertex in $\Gamma$ 9 is adjacent to every vertex in $D$ 0. The vertices of $D$ 1 can be assigned levels according to the number of in-neighbors in $D$ 2, and the main theorem states that all vertices in $D$ 3, except possibly one, lie in two consecutive levels (Hell et al., 23 Sep 2025).

Balanced digraphs are the class where the definition becomes most transparent. A balanced digraph is an irreflexive digraph in which every cycle has the same number of forward and backward arcs; equivalently, the vertex set can be partitioned into $D$ 4 so that every arc goes from $D$ 5 to $D$ 6. If $D$ 7 denotes the underlying bipartite graph between $D$ 8 and $D$ 9, then the digraph is strongly chordal if and only if each $M(D)$ 0 is a chordal bigraph. Equivalently, a balanced digraph is strongly chordal exactly when it contains no induced fence, where a fence is an oriented even cycle of length greater than $M(D)$ 1 with no directed path of length $M(D)$ 2. An immediate corollary is that every oriented tree is strongly chordal (Hell et al., 2019, Hell et al., 23 Sep 2025).

5. Recognition algorithms and domination theory

Recognition is polynomial in several important settings, but not in full generality. The foundational study established polynomial-time recognition for symmetric digraphs, tournaments with possible loops, and balanced digraphs, together with forbidden induced subgraph characterizations in each of these cases (Hell et al., 2019). Later work extended this program to additional families and supplied $M(D)$ 3-time certifying algorithms for irreflexive tournaments, tournaments with possible loops, reflexive multipartite tournaments, and several related classes (Hell et al., 23 Sep 2025).

The algorithmic viewpoint is inseparable from the ordering viewpoint. In the symmetric case, the equivalence between strong chordality, total balance, simple vertices in every induced subgraph, and simple orderings gives a direct elimination-based recognition mechanism. In tournaments and multipartite tournaments, recognition uses strong components, degree-sequence tests, explicit obstruction families, and constructive production of a $M(D)$ 4-free ordering when the input belongs to the class (Hell et al., 2019, Hell et al., 23 Sep 2025).

A second major algorithmic theme is domination. For strongly chordal graphs with possible loops, a linear-time algorithm computes a minimum general dominating set when a strong ordering is available. The algorithm maintains a set $M(D)$ 5 of vertices with pairwise disjoint neighborhoods, a set $M(D)$ 6 of chosen dominators, and a temporary label $M(D)$ 7: repeatedly one chooses the first vertex $M(D)$ 8 not labeled $M(D)$ 9, lets $v_1,\dots,v_n$ 0 be the last neighbor of $v_1,\dots,v_n$ 1, labels $v_1,\dots,v_n$ 2 with $v_1,\dots,v_n$ 3, labels $v_1,\dots,v_n$ 4 with $v_1,\dots,v_n$ 5, and labels all neighbors of $v_1,\dots,v_n$ 6 with $v_1,\dots,v_n$ 7. The invariants $v_1,\dots,v_n$ 8 and pairwise disjointness of neighborhoods of vertices in $v_1,\dots,v_n$ 9 imply optimality. This extends and unifies Farber’s algorithm for minimum dominating set in reflexive strongly chordal graphs and the Damaschke–Mueller–Kratsch algorithm for minimum total dominating set in chordal bipartite graphs (Hell et al., 2019).

For strongly chordal digraphs, later work proves a min–max theorem of the same flavor. The paper states that in a strongly chordal digraph the minimum size of a total dominating set equals the maximum number of disjoint in-neighborhoods, and this number can be calculated in linear time given a $V(D)$ 0-free ordering. In the detailed development, domination is formulated so that a looped vertex may dominate itself while a loopless vertex cannot, thereby unifying domination and total domination. The greedy procedure again relies on a strong ordering: select the first unlabeled vertex $V(D)$ 1, choose the last in-neighbor $V(D)$ 2 of $V(D)$ 3, label $V(D)$ 4 by $V(D)$ 5, $V(D)$ 6 by $V(D)$ 7, and all out-neighbors of $V(D)$ 8 by $V(D)$ 9. The correctness argument is that two $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$00-vertices cannot have intersecting in-neighborhoods without creating a forbidden $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$01-submatrix (Hell et al., 23 Sep 2025).

Strongly chordal digraphs are closely related to, but distinct from, several neighboring notions. One of these is the class of strict chordal digraphs studied by Hell and Hernández-Cruz. A later signed-graph translation states that strict chordal digraphs are exactly the digraphs whose associated signed graphs are chordal, where a symmetric arc pair becomes a positive edge and a non-symmetric arc adjacency becomes a negative edge. This yields a different but structurally parallel framework based on signed simplicial vertices, signed simplicial edges, separator structure, and lollipop decompositions (Huang et al., 16 Dec 2025).

Another neighboring notion is the cycle-based “chordal directed graph” condition. In that setting, a digraph is chordal directed when it has no induced directed cycle of length at least $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$02. The negative dichromatic result shows that even the class

$\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$03

has unbounded dichromatic number: for every $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$04, there exists $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$05 with $\Gamma=\begin{bmatrix}1&1\[2pt]1&0\end{bmatrix}.$06. A plausible implication is that strong chordality should be understood primarily as a matrix-orderability property rather than as a mere strengthening of directed cycle restrictions (Aboulker et al., 2022).

There is also a broader algorithmic chordality literature that does not directly concern strongly chordal digraphs. For example, the Edmonds–Giles conjecture on packing dijoins has been proved for weighted digraphs whose underlying undirected graph is chordal, together with a strongly polynomial-time construction. That work does not state a theorem for strongly chordal digraphs, but it indicates that ordinary chordality of the underlying graph already supports strong min–max structure in some directed optimization problems (Cornuéjols et al., 19 Jan 2025).

The main open direction emphasized in the strong-chordality literature is general recognition. The foundational paper states that in general it is not clear whether strongly chordal digraphs can be recognized in polynomial time, and the later survey of specialized classes again leaves open the general recognition problem for arbitrary digraphs with possible loops (Hell et al., 2019, Hell et al., 23 Sep 2025). At present, the theory is therefore most complete in structured families—symmetric graphs with loops, tournaments, multipartite tournaments, balanced digraphs, and closely related classes—where strong orderings, obstruction sets, and elimination arguments can be made explicit.